AI Proctoring Failure Forces 58,000 Mexican University Candidates to Retake Exams
The National Autonomous University of Mexico (UNAM) ordered approximately 58,000 candidates to retake their entrance exams in person after a new AI-powered remote proctoring system failed to prevent widespread cheating. The failure was exposed when the percentage of high-scoring candidates spiked dramatically compared to previous years. This incident highlights the limitations of current AI proctoring technologies and the significant challenges of securing high-stakes remote examinations at scale. It serves as a cautionary tale for educational institutions relying solely on automated systems to maintain academic integrity. Despite using LockDown Browser and AI to monitor audio, video, and screen activity, the percentage of students scoring over 100 points rose from 3.5% in previous years to 16.3%. Consequently, an expert committee determined that a controlled, in-person exam was the only way to ensure fair results.
## BACKGROUND
Remote proctoring tools like LockDown Browser are designed to secure online exams by restricting computer functionalities such as copy-pasting, screen recording, and opening other applications. AI proctoring adds another layer by using computer vision and audio analysis to detect suspicious behaviors, such as unauthorized devices or secondary users in the room.